30
Stata Technical Bulletin
STB-57
logit (N=753): Factor Change in Odds
Odds of: inLF vs NotInLF
|
Ifp I |
b |
z |
P>∣z∣ |
e^b |
e^bStdX |
SDofX |
|
kε I |
-1.46291 |
-7.426 |
0.000 |
0.2316 |
0.4646 |
— 0.5240 |
|
k618 I |
-0.06457 |
-0.950 |
0.342 |
0.9375 |
0.9183 |
1.3199 |
|
age I |
-0.06287 |
-4.918 |
0.000 |
0.9391 |
0.6020 |
8.0726 |
|
wc I |
0.80727 |
3.510 |
0.000 |
2.2418 |
1.4381 |
0.4500 |
|
he I |
0.11173 |
0.542 |
0.588 |
1.1182 |
1.0561 |
0.4885 |
|
Iwg I |
0.60469 |
4.009 |
0.000 |
1.8307 |
1.4266 |
0.5876 |
|
inc I |
-0.03445 |
-4.196 |
0.000 |
0.9661 |
0.6698 |
11.6348 |
|
— b = P>∣z∣ = |
raw coefficient |
b=0 |
— | |||
e^b = exp(b) = factor change in odds for unit increase in X
e^bStdX = exp(b*SD of X) = change in odds for SD increase in X
SDofX = standard deviation of X
Alternatively, a user might want to interpret the coefficients in terms of their effect on the latent y* that underlies the observed
variable y. Note that to use Iistcoef to compute standardized coefficients does not require that the model be reestimated.
. Iistcoef, std help
logit (N=753): Unstandardized and Standardized Estimates
Observed SD: .49562951
Latent SD: 2.0500391
Odds of: inLF vs NotInLF
|
— Ifp I |
b |
z |
P>∣z∣ |
bStdX |
bStdY |
bStdXY |
— SDofX |
|
k5 I |
-1.46291 |
-7.426 |
0.000 |
-0.7665 |
-0.7136 |
-0.3739 |
— 0.5240 |
|
k618 I |
-0.06457 |
-0.950 |
0.342 |
-0.0852 |
-0.0315 |
-0.0416 |
1.3199 |
|
age I |
-0.06287 |
-4.918 |
0.000 |
-0.5075 |
-0.0307 |
-0.2476 |
8.0726 |
|
we I |
0.80727 |
3.510 |
0.000 |
0.3633 |
0.3938 |
0.1772 |
0.4500 |
|
he I |
0.11173 |
0.542 |
0.588 |
0.0546 |
0.0545 |
0.0266 |
0.4885 |
|
Iwg I |
0.60469 |
4.009 |
0.000 |
0.3553 |
0.2950 |
0.1733 |
0.5876 |
|
inc I |
-0.03445 |
-4.196 |
0.000 |
-0.4008 |
-0.0168 |
-0.1955 |
11.6348 |
b = raw coefficient
z = z-score for test of b=0
P>∣z∣ = р-value for z-test
bStdX = х-standardized coefficient
bStdY = у-standardized coefficient
bStdXY = fully standardized coefficient
SDofX = standard deviation of X
Example with mlogit
A key to fully interpreting the multinomial logit model is to consider all contrasts among the outcome categories. The
standard output from mlogit provides contrast between all outcomes and the category specified by basecategory. Here we
specify the rrr option in order to obtain “relative risk ratios” which are also known as factor change coefficients.
|
. mlogit occ white ed exper, basecategory(1) rrr nolog | ||||||
|
Multinomial regression Log likelihood = -426.80048 |
Number of obs = Prob > chi2 = Pseudo R2 = |
337 | ||||
|
— occ I |
RRR |
Std. Err. |
z |
P>∣z∣ |
[957. Conf. |
— Interval] |
|
BlueCol I |
3.443553 |
2.494631 |
1.707 -0.972 0.271 |
0.088 0.331 0.786 |
.8324658 .7408981 .9710484 |
— 14.2445 1.106324 1.039585 |
|
Craft I white I |
1.603748 1.098357 1.028071 |
.9691607 .1071502 .0171417 |
0.782 0.962 1.660 |
0.434 0.336 0.097 |
.4906218 .9072032 .9950164 |
— 5.242345 1.329788 1.062223 — |
More intriguing information
1. The Macroeconomic Determinants of Volatility in Precious Metals Markets2. Bargaining Power and Equilibrium Consumption
3. The name is absent
4. Iconic memory or icon?
5. Impact of Ethanol Production on U.S. and Regional Gasoline Prices and On the Profitability of U.S. Oil Refinery Industry
6. The name is absent
7. The name is absent
8. Behaviour-based Knowledge Systems: An Epigenetic Path from Behaviour to Knowledge
9. LIMITS OF PUBLIC POLICY EDUCATION
10. Imitation in location choice